Page 42 - ITU Journal Future and evolving technologies – Volume 2 (2021), Issue 2
P. 42
ITU Journal on Future and Evolving Technologies, Volume 2 (2021), Issue 2
mon RB with similar output power levels (low PAPR is en‑ ment and content delivery is performed in every time slot,
sured) without loss of generality we may assume the use without any collisions and threshold constraint. More‑
of OFDMA which enables the use of subcarriers by differ‑ over for all igures, the gap between proposed procedures
ent devices [51]. and the corresponding benchmark schemes presents the
gain provided by clustering and of loading.
7. SIMULATION RESULTS
In this section performance evaluation of proposed In a mild condition such as, low user density, low SINR
ABD2D and DBD2D procedures are presented. Number threshold, and long Messaging Periof (MP) duration, of‑
of required control messages for unit of loading and aver‑ loading amounts provided by ABD2D and DBD2D are
age amount of of loading (number of contents delivered expected to convergence to DBCA + HCA’s performance.
using D2D transmissions) are considered as key perfor‑ Likewise, convergence to MPCC is expected for bench‑
mance metrics and investigated against a varying number mark schemes. Accordingly in the simulations, variables
of HUs, OUs, SINR threshold and allocated MP length for are de ined in a wide‑scale to observe results for both
discovery and advertisement. Simulations are performed challenging and moderate conditions.
in MATLAB, for the parameters in Table 1.
Table 1 – Simulation parameters MP duration is de ined between 1‑20 ms where a 1 ms
long MP provides only 10 RBs but very low latency for el‑
Parameter Value igible users. On the other hand, as mentioned before, a 1
Number of OUs [100,250,400,550,700] ms extension of MP duration results in a 4 ms extension
Number of HUs [30,45,60,75,90] for AZ and longer MP durations are not practical in terms
MP duration [1,2,3,4,5,10,15,20] ms of latency. Moreover, effects of the number of HUs and
SINR threshold [‑10,‑5,0,5,10,15,20] dB OUs are investigated for a 3 and 7 times increase, respec‑
Zipf dist. skewness ( ) 0.6 tively. Lastly, the entire scale for the SINR threshold in‑
Number of contents ( ) 20 cludes QPSK (Quadrature Phase Shift Keying) to 64‑QAM
Cache capacity ( ) 4 units (Quadrature Amplitude Modulation) values for different
Content size ( ) 1 unit coding rates.
Number of random seeds 100
Simulation duration 100×period
7
Period duration 2 s ABD2D-BM
Radius of cellular area 250 m 6.5 ABD2D
DBD2D-BM
Mobility model RDMM 6 DBD2D
Max speed of users 3 m/s 5.5
Transmission power ( ) 200 mW, 23 dBm 5
Height of users ℎ , ℎ 1.5 m Number of Beacons for Unit Offloading 4.5
Operating frequency 2 GHz 4
ABD2D‑BM and DBD2D‑BM are different forms of pro‑ 3.5
posed procedures and presented for benchmarking. In 3
these procedures, DBCA clustering and HCA cache place‑ 2.5
ment algorithms are not applied. All HUs act as IHUs; 2
cache most popular content, broadcast advertisement 1 2 3 4 5 10 15 20
MP Duration (ms)
beacons in ABD2D‑BM and listen to discovery channels
individually in DBD2D‑BM. Fig. 7 – Mean number of beacons for unit of loading for varying MP du‑
ration
Another benchmark scheme is DBCA+HCA. It excludes
proposed discovery and advertisement procedures. In Fig. 7 shows the average number of transmitted bea‑
other words, an OU can reach of loaded content without con messages per successful content of loading as a func‑
any collisions, if it has a link with the corresponding HU tion of MP duration. Especially for the DBD2D proce‑
that ful ills the minimum SINR requirement. Lastly, the dure, it is clear that a long MP duration leads to success‑
simplest benchmark scheme is the Most Popular Content ful discovery attempts and fewer beacon messages are re‑
Caching (MPCC ). MPCC is exempt from clustering, hierar‑ quired for successful of loading. On the other hand, for
chical cache placement, advertisement and discovery pro‑ the ABD2D procedure the number of beacons tend to stay
cedures. HUs do not form clusters and all of them act as stable, though there is a 20‑fold increase of the MP dura‑
IHUs and cache the most popular content. tion. The reason for such a tendency is the adequately al‑
located RBs. A 1 ms long MP provides 10 RBs and thanks
Benchmark DBCA+MPCC scheme provides the upper to clustering it is enough for 60 HUs, since only CHs trans‑
bound for of loading, where the clustering, content place‑ mit the advertisement beacons.
28 © International Telecommunication Union, 2021